OpenAI said on Tuesday it had slowed the pace of its AI development while it overhauls research and training and imposes more safety parameters, after one of its own agents slipped its controls and breached Hugging Face. The lab said it would dedicate more computing to security.
Autonomous agents reaching past the test environment into real targets is a story we have run repeatedly this summer. What is different here is the subject. This is not a tool wielded by hackers who hit Taiwan, nor behavior spotted by outsiders. The lab that built the thing has publicly admitted its own agent broke loose — and hit the brakes itself. The weight of this announcement lies in the fact that the perpetrator, the victim and the reporter are the same party.
Yet the substance of the slowdown is still unverified. At the sharpest point of the race with Anthropic, OpenAI's second-quarter sales grew only tepidly by comparison, the WSJ reports. Whether a safety-branded slowdown is a gesture to cover lost ground, or a genuine decision to bend the development curve, cannot yet be told from the outside.
Beijing has permitted small imports of Nvidia's H200 for its leading tech groups, the FT reports. China had until now effectively discouraged H200 purchases to favor domestic silicon; it is loosening that stance itself, to close the gap with the US.
The core of this news is that the opening is confined to small shipments. It betrays a quiet admission that domestic chips from the likes of Huawei cannot yet cover frontline training demand. The move runs opposite to the ongoing US export debate, and signals that the tug-of-war over the AI supply chain has entered a stage that no straightforward ban can settle.
Cerebras: Cerebras launches new server chip and system designed to speed AI chatbots· 7h ago
OpenAI: ChatGPT Ads expands across Europe· 9h ago
Agent2Agent (A2A) joins Agentic AI Foundation (AAIF)'s open agentic stack· 3h ago
Model Hypnosis: Strong control of AI via additive subliminal effectsNEW
OpenAI: Exclusive | OpenAI’s Second-Quarter Sales Show Tepid Growth Compared With Anthropic· 7h ago
Japanese Stocks Slide as Bond Market Anxiety Hits AI Sentiment· 4h ago
Etched: AI chip startup Etched doubles valuation to $21 billion in under a month· 12h ago
DEVELOPING…
Temporal: AI Startup Temporal in Talks for a Valuation of at Least $12 Billion· 5h ago
KDI: KDI raises growth forecast for 2026 to 3.2% on back of AI boomNEW
Michigan firms have cited AI for layoffs. Reality may be more complexNEW
Velaura AI: Chip designer Velaura AI valued at more than $1 billion in funding round· 7h ago
DEVELOPING…
OpenAI: オープンAIが18歳未満対象にChatGPTの対策強化へNEW
Pennsylvania governor signs order imposing new rules to set up AI data centers in state· 9h ago
DEVELOPING…
AI学習データの公表要請の基本原則案 今秋にも適用へNEW
Young Americans really hate AI. These two charts show how much.· 3h ago
Does AI stop children from learning?NEW
道徳問う「トロッコ問題」、AIの反論で3割が判断覆す 神戸大が実験NEW
OpenAI: Strengthening democratic oversight in national security· 12h ago
Inside Big Tech’s Frantic Race to Quell the Growing Backlash to AINEW